检索结果-内蒙古大学图书馆

Path-Gradient Estimators for Continuous Normalizing Flows

学校读者我要写书评

暂无评论

arXiv 2022年

Recent work has established a path-gradient estimator for simple variational Gaussian distributions and has argued that the path-gradient is particularly beneficial in the regime in which the variational distribution approaches the exact target distribution. In many applications, this regime can however not be reached by a simple Gaussian variational distribution. In this work, we overcome this crucial limitation by proposing a path-gradient estimator for the considerably more expressive variational family of continuous normalizing flows. We outline an efficient algorithm to calculate this estimator and establish its superior performance empirically. © 2022, CC BY.

关键词：

Evaluating Posterior Distributions by Selectively Breeding Prior Samples

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Shalizi, Cosma Rohilla Department of Statistics and Data Science and of Machine Learning Carnegie Mellon University PittsburghPA15213 United States The Santa Fe Institute 1399 Hyde Park Road Santa FeNM87501 United States

Using Markov chain Monte Carlo to sample from posterior distributions was the key innovation which made Bayesian data analysis practical. Notoriously, however, MCMC is hard to tune, hard to diagnose, and hard to parallelize. This pedagogical note explores variants on a universal non-Markov-chain Monte Carlo scheme for sampling from posterior distributions. The basic idea is to draw parameter values from the prior distributions, evaluate the likelihood of each draw, and then copy that draw a number of times proportional to its likelihood. The distribution after copying is an approximation to the posterior which becomes exact as the number of initial samples goes to infinity;the convergence of the approximation is easily analyzed, and is uniform over Glivenko-Cantelli classes. While not entirely practical, the schemes are straightforward to implement (a few lines of R), easily parallelized, and require no rejection, burn-in, convergence diagnostics, or tuning of any control settings. I provide references to the prior art which deals with some of the practical obstacles, at some cost in computational and analytical simplicity. Copyright © 2022, The Authors. All rights reserved.

关键词： Markov processes

Flow-Based Sampling for Entanglement Entropy and the machine learning of Defects

学校读者我要写书评

暂无评论

Physical Review Letters 2025年第15期134卷 151601-151601页

作者： Andrea Bulgarelli Elia Cellini Karl Jansen Stefan Kühn Alessandro Nada Shinichi Nakajima Kim A. Nicoli Marco Panero Department of Physics University of Turin and INFN Turin unit Via Pietro Giuria 1 I-10125 Turin Italy Computation-Based Science and Technology Research Center The Cyprus Institute Nicosia Cyprus Deutsches Elektronen-Synchrotron DESY Zeuthen Germany Berlin Institute for the Foundations of Learning and Data (BIFOLD) Berlin Germany Machine Learning Group Technische Universität Berlin Berlin Germany RIKEN Center for AIP Tokyo Japan Transdisciplinary Research Area (TRA) Matter University of Bonn Germany Helmholtz Institute for Radiation and Nuclear Physics (HISKP) Bonn Germany Department of Physics and Helsinki Institute of Physics PL 64 FIN-00014 University of Helsinki Finland

We introduce a novel technique to numerically calculate Rényi entanglement entropies in lattice quantum field theory using generative models. We describe how flow-based approaches can be combined with the replica trick using a custom neural-network architecture around a lattice defect connecting two replicas. Numerical tests for the ϕ4 scalar field theory in two and three dimensions demonstrate that our technique outperforms state-of-the-art Monte Carlo calculations, and exhibit a promising scaling with the defect size.

关键词： Lattice theory

Secure Device on boarding in IoT Networks

学校读者我要写书评

暂无评论

Secure Device on boarding in IoT Networks

International Conference on science Technology Engineering and Management (ICONSTEM)

作者： P. Sathyaraj Shankar Nayak Bhukya S. Rukmani Devi Chetan Umadi A. Ajina Rajendiran M Department of Electronics and Communication Engineering RMK College of Engineering and Technology Puduvoyal Tamil Nadu India Department of Computer Science Engineering (Data Science) CMR Technical Campus Hyderabad Telangana India Department of Computer Science Saveetha College of Liberal Arts and Sciences SIMATS Deemed to be University Chennai Tamil Nadu India Department of Electronics& Telecommunication Engineering Dayananda Sagar College of Engineering Bengaluru Karnataka India Department of Artificial Intelligence and Machine Learning M S Ramaiah Institute of Technology Bangalore India Department of Computer Science and Engineering Panimalar Engineering College Chennai Tamil Nadu India

ISBN: (数字)9798350365092

ISBN: (纸本)9798350365108

The fast spread of Internet of Things (IoT) gadgets has led to unprecedented number of interconnected systems offering many applications and services. Diverse elements in IoT networks increase security vulnerabilities, especially when adding new devices. IoT device onboarding secures device integration, authentication, and network connection authorization. This research shows the hazards and limitations of secure IoT device connections. Research examines onboarding protocols' merits and downsides. Bootstrapping, mutual authentication, and key establishment protect the IoT device-network connection in the proposed architecture. Overcoming scalability and resource constraints makes onboarding several IoT devices straightforward. We also analyze IoT network security with PKI and certificate-based authentication. Safe communication pathways and ongoing observation and system adjustments to address new security issues are discussed in this essay. Researchers test the technology in IoT environments. The framework handles onboarding in this case. The method safeguards onboarding and IoT ecosystem data, research demonstrate. A detailed device onboarding approach increases IoT security talks. The framework fixes onboarding issues and vulnerabilities for secure IoT networks.

关键词： Protocols Scalability Ecosystems Authentication Interconnected systems Network security Mathematics

Deep Reinforcement learning for Beam Angle Optimization of Intensity-Modulated Radiation Therapy

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Bao, Peng Wang, Gong Yang, Ruijie Dong, Bin The Center for Data Science Peking University the Department of Radiation Oncology Peking University Third Hospital Beijing China The Department of Radiation Oncology Peking University Third Hospital Beijing China Beijing International Center for Mathematical Research Center for Machine Learning Research National Biomedical Imaging Center Peking University Beijing China

Objective: Intensity-modulated radiation therapy (IMRT) beam angle optimization (BAO) is a challenging combinatorial optimization problem that is NP-hard. In this study, we aim to develop a personalized BAO algorithm for IMRT that improves the quality of the final treatment. Methods: To improve the quality of IMRT treatment planning, we propose a deep reinforcement learning (DRL)based approach for IMRT BAO. We consider the task as a sequential decision-making problem and formulate it as a Markov Decision Process. To facilitate the training process, a 3D-Unet is designed to predict the dose distribution for the different number of beam angles, ranging from 1 to 9, to simulate the IMRT environment. By leveraging the simulation model, double deep-Q network (DDQN) and proximal policy optimization (PPO) are used to train agents to select the personalized beam angle sequentially within a few seconds. Results: The treatment plans with beam angles selected by DRL outperform those with clinically used evenly distributed beam angles. For DDQN, the overall average improvement of the CIs is 0.027, 0.032, and 0.03 for 5, 7, and 9 beam angles respectively. For PPO, the overall average improvement of CIs is 0.045, 0.051, and 0.025 for 5, 7, and 9 beam angles respectively. Conclusion: The proposed DRL-based beam angle selection strategy can generate personalized beam angles within a few seconds, and the resulting treatment plan is superior to that obtained using evenly distributed angles. Significance: A fast and automated personalized beam angle selection approach is been proposed for IMRT BAO. © 2023, CC BY.

关键词： Reinforcement learning

Facilitated machine learning for image-based fruit quality assessment

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Knott, Manuel Perez-Cruz, Fernando Defraeye, Thijs Empa Swiss Federal Laboratories for Materials Science Technology Laboratory for Biomimetic Membranes and Textiles St. Gallen Switzerland Swiss Data Science Center ETH Zurich and EPFL Zurich Switzerland Institute for Machine Learning Department of Computer Science ETH Zurich Switzerland

Image-based machine learning models can be used to make the sorting and grading of agricultural products more efficient. In many regions, implementing such systems can be difficult due to the lack of centralization and automation of postharvest supply chains. Stakeholders are often too small to specialize in machine learning, and large training data sets are unavailable. We propose a machine learning procedure for images based on pre-trained Vision Transformers. It is easier to implement than the current standard approach of training Convolutional Neural Networks (CNNs) as we do not (re-)train deep neural networks. We evaluate our approach based on two data sets for apple defect detection and banana ripeness estimation. Our model achieves a competitive classification accuracy equal to or less than one percent below the best-performing CNN. At the same time, it requires three times fewer training samples to achieve a 90% accuracy. © 2022, CC BY.

关键词： Computer vision

Precise Asymptotics of Bagging Regularized M-estimators

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Koriyama, Takuya Patil, Pratik Du, Jin-Hong Tan, Kai Bellec, Pierre C. Booth School of Business The University of Chicago ChicagoIL60637 United States Department of Statistics University of California BerkeleyCA94720 United States Department of Statistics and Data Science Carnegie Mellon University PittsburghPA15213 United States Machine Learning Department Carnegie Mellon University PittsburghPA15213 United States Department of Statistics Rutgers University New BrunswickNJ08854 United States

We characterize the squared prediction risk of ensemble estimators obtained through subagging (subsample bootstrap aggregating) regularized M-estimators and construct a consistent estimator for the risk. Specifically, we consider a heterogeneous collection of M ≥ 1 regularized M-estimators, each trained with (possibly different) subsample sizes, convex differentiable losses, and convex regularizers. We operate under the proportional asymptotics regime, where the sample size n, feature size p, and subsample sizes km for m ∈ [M] all diverge with fixed limiting ratios n/p and km/n. Key to our analysis is a new result on the joint asymptotic behavior of correlations between the estimator and residual errors on overlapping subsamples, governed through a (provably) contractible nonlinear system of equations. Of independent interest, we also establish convergence of trace functionals related to degrees of freedom in the non-ensemble setting (with M = 1) along the way, extending previously known cases for square loss and ridge, lasso regularizers. When specialized to homogeneous ensembles trained with a common loss, regularizer, and subsample size, the risk characterization sheds some light on the implicit regularization effect due to the ensemble and subsample sizes (M, k). For any ensemble size M, optimally tuning subsample size yields sample-wise monotonic risk. For the full-ensemble estimator (when M → ∞), the optimal subsample size k* tends to be in the overparameterized regime (k* ≤ min{n, p}), when explicit regularization is vanishing. Finally, joint optimization of subsample size, ensemble size, and regularization can significantly outperform regularizer optimization alone on the full data (without any subagging). Copyright © 2024, The Authors. All rights reserved.

关键词： Nonlinear equations

AI-Optimized Placement and Routing for PCB Design

学校读者我要写书评

暂无评论

AI-Optimized Placement and Routing for PCB Design

International Conference on science Technology Engineering and Management (ICONSTEM)

作者： P. Sathyaraj S. Arulkumar A. Ajina Kedir Beshir L. Umasankar A.N. Arularasan Department of Electronics and Communication Engineering RMK College of Engineering and Technology Puduvoyal Tamil Nadu India Department of Electrical and Computer Engineering Wachemo University Ethiopia Department of Artificial Intelligence and Machine Learning M S Ramaiah Institute of Technology Bangalore India Department of Electrical and Electronics Engineering S.A. Engineering College Tiruvallur Tamil Nadu India Department of Artificial Intelligence and Data Science Panimalar Engineering College Chennai Tamil Nadu India

ISBN: (数字)9798350365092

ISBN: (纸本)9798350365108

Rapid electronic device development requires more complicated and densely packed PCB designs. These systems need properly placed and connected electrical components for best performance and reliability. Complexity and time restrictions are rendering manual methods ineffective. Therefore, AI-based PCB insertion and routing automation methods have been developed. This research suggests using AI to optimize PCB component arrangement and routing. The system uses machine learning and optimization to improve design efficiency. The AI system creates layouts that match performance requirements, decrease signal interference, and meet manufacturing limits using previous design data and user preferences. A learning module that updates its knowledge base, an algorithm that chooses between design objectives, and real-time user input are examples. Signal integrity, thermal issues, and manufacturability are considered while modifying design requirements. Evaluation of the suggested methodology shows considerable gains in design efficiency, time-to-market, and electronic system performance. This research improves AI-automated PCB design, enabling more complex and effective electrical systems. AI in PCB design will revolutionize the industry by solving complicated design problems and speeding up the development of cutting-edge electronics.

关键词： machine learning algorithms System performance machine learning Routing Rendering (computer graphics) Real-time systems Reliability

Advancing Alzheimer's Nodule Detection Through a Comprehensive Multi-Scene Deep learning Framework

学校读者我要写书评

暂无评论

Advancing Alzheimer's Nodule Detection Through a Comprehensi...

International Conference on Advanced Computing and Communication Systems (ICACCS)

作者： C. Premkumar R. Rajeshwari R. Remya M. S. Cholaathiraj A. Reethika S. Bharathidasan Department of Artificial Intelligence and Data Science Bannari Amman Institute of Technology Sathyamangalam India Department of Computer Science and Engineering CMR Institute of Technology Bengaluru India Department of Electronics and Communication Engineering Vel Tech Rangarajan Dr Sagunthala R&D institute of science and Technology Chennai India Department of Artificial Intelligence and Machine Learning Bannari Amman Institute of Technology Sathyamangalam India Department of Electronics and Communication Engineering Sri Ramakrishna Engineering College Coimbatore India Department of Electronics and Communication Engineering Erode Sengunthar Engineering College Erode India

ISBN: (数字)9798350384369

ISBN: (纸本)9798350384376

Determining the precise location of Alzheimer's nodules is essential for estimating the risk of brain cancer. Conventional CAD modules, including MRI, PET, and CT, struggle with feature extraction and segmentation due to time limits and complexity. This study proposes an efficient brain nodule detection technique based on the Multi-Scene Deep learning Framework (MSDLF) that makes use of the vesselness filter. A four-channel CNN model is developed using information from two image sequences to enhance radiologists' ability to recognise four-stage nodules. This adaptable model may be used in two different classes. The outcomes demonstrate how well the MSDLF performs while processing substantial volumes of image data for brain nodule recognition, improving with accuracy 98.86% and 98.45% of sensitivity by using decreasing false positives.

关键词： Deep learning Training Solid modeling Accuracy Sensitivity Image recognition Magnetic resonance imaging Brain modeling Classification algorithms Alzheimer's disease